Abstract: In the United States, law enforcement agencies kill three people per day. Yet, only around half of these deaths are recorded as police killings in official statistics. This has serious consequences for police accountability and citizens' trust in public institutions. In the US, all unnatural deaths must be investigated by either a coroner or medical examiner depending on the state. Usually, coroners are elected officials without a medical background, while medical examiners are appointed and have medical qualifications. This paper studies the impact of the medicolegal death investigation system on the (under)reporting of police killings. Using historical data, I construct a panel detailing the characteristics of the death investigation systems across states from 1968 to 2020. I exploit time and state variation using a difference-in-difference analysis to estimate the impact of having a medical examiner (vis-à-vis coroners) on (i) the official number of police killings, (ii) number of police killings based on detailed geo-coded crowdsource data, and (iii) underreporting of police killings. Preliminary findings show no evidence in favor of a system per se. Elections vis-a-vis appointments do not impact reporting. Moreover, underreporting is not statistically different during electoral years. Also, requiring medical expertise does not affect under-reporting. However, the independence of the medico-legal investigation office and the local law enforcement agency matters. Centralized medico-legal investigations offices reduce under-reporting. Medico-legal offices that are joined with law enforcement agencies have higher under-reporting.